Ashish Verma-1011015
12/1/2011
Note: Conditioning of data is being done before analysis. Gaps or missing data points are estimated by linear
correlation.
O A REPOR
QUESTION 1: COMPUTE THE DAILY LOG RETU RNS OF THE FOUR EXCHANGE RATES.
Euro
Pound
Dollar
Log-returns
Log-returns
.04
.04
.02
.02
Yen
LOGRETURN_DOLLAR
LOGRETURN_YEN
.03
.05
.02
.04
.03
.01
.02
.00
.00
.00
.01
-.02
-.01
-.02
-.04
.00
-.02
-.01
-.03
-.02
-.04
-.06
01
-.06
01
02
03
04
05
06
07
08
09
10
02
03
04
05
06
07
08
09
10
11
-.03
-.04
11
01
02
03
04
05
06
07
08
09
10
11
-.04
01
02
03
04
05
06
07
08
09
10
11
Euro
. 16
.1
.7798
-.19
.667
-.668
7.819691
Pound
4.7E-
. 7E-
.1 94
-.189
. 94
-. 7 6
8.66
Jarque-Bera
Probabili
7 1.8
um
um q. Dev.
Observaions
Mean
Median
Maximum
Minimum
d. Dev.
ewness
Kurosis
X
X
X
G
Yen
3.39E-05
0.000173
-1.37E-06
4.41E-05
0.010815
0.045163
-0.013057
-0.037612
0.001393
0.006489
0.453996
0.385977
12.28561
8.217757
79.66
10191.76
3257.364
0.000000
0.000000
.6711
.9 9
.1 4
.161
0.095214
0.487059
0.005448
0.118283
81
81
2810
2810
Returns are negatively skewed for Euro, Pound but positively skewed for dollar and yen
Returns are not normal for all currencies and are leptokurtic or fat-tailed
Conditional expected return are close to zero for all currencies
|
|
G
Currenc
Euro
Pound
Dollar
Yen
-53.6678
0.0001
-51.2514
0.0001
-.14 9
0.0001
-52.1018
0.0001
At . confidence level, we reject the null hypothesis; log returns of all series do not have a unit root and
are stationary.
Correlogram of log returns
Euro
Gollar
^
As seen in the ACF/PCF plot of all series, it does not have very high degree of auto-correlation whereas a nonstationary series has high degree of auto-correlation decreasing slowly with increasing lags. Returns series
does show little serial autocorrelation which is also one of the stylized facts.
Euro
Pound
700
700
Series: LOG_RETURNS
Sample 1/02/2001 11/16/2011
Observations 2810
600
Series: LOG_RETURNS
Sample 1/02/2001 11/16/2011
Observations 2810
600
500
500
Mean
Median
Maxim um
Minim um
Std. Dev.
Skewness
Kurtosis
400
300
200
0.000216
0.000123
0.037798
-0.051593
0.005667
-0.066583
7.819691
400
300
200
0
-0.015
0
-0.050
-0.025
0.000
4.70e-05
2.27e-05
0.012940
-0.018095
0.002394
-0.325726
8.656333
Jarque-Bera 3795.665
Probability
0.000000
100
Jarque-Bera 2721.855
Probability 0.000000
100
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
-0.010
-0.005
0.000
0.005
0.010
0.025
Dollar
Yen
700
1,200
Series: LOGRETURN_DOLLAR
Sample 1/02/2001 11/16/2011
Observations 2810
1,000
Series: LOGRETURN_YEN
Sample 1/02/2001 11/16/2011
Observations 2810
600
500
Mean
Median
Maxim um
Minim um
Std. Dev.
Skewness
Kurtosis
800
600
400
200
7.80e-05
-3.15e-06
0.024903
-0.030065
0.003207
0.453996
12.28561
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
400
300
200
Jarque-Bera 3257.364
Probability 0.000000
100
Jarque-Bera 10191.76
Probability 0.000000
0
-0.025
0
-0.03
-0.02
-0.01
-0.00
0.01
0.000173
4.41e-05
0.045163
-0.037612
0.006489
0.385977
8.217757
0.000
0.025
0.02
Return series for all currencies are not normal and is leptokurtic (p-value of Jarque-Bera is less than for all
series thus the null hypothesis (return series is normal) is rejected)
Q-: BUILD AN AR MODEL FOR EACH SERIES AND CHECK THE FITTED MODEL.
Oied AR Models
Euro
Pound
0.000215
2.159146
0.0309
!"#
!$#
-0.03424
-1.81336
0.0699
!$#
!""#
-0.03958
-2.09751
0.036
-7.51176
0.03686
1.955399
-0.04857
0.0506
-2.57808
0.01
-9.23493
Dollar
!"#
!$#
!%#
!&#
Yen
.6 1
. 67
.891
-.66 4
3.304659
1.767634
4.717954
-3.49361
Euro
Gollar
.1 AR(9)
-.1 -1.666
.77
AR(10) .7489 .98 16
.
-8.686
.98
.1
-7. 461
Correlogram of Residuals
^
Although Residuals showed very little correlation in Fitted AR models in all 4 cases, but the residuals
in all fours were not normal. JB test on residuals showed that residuals did not followed normality
assumption. They were leptokurtic in nature.
QUESTION 6: BUILD AN MA MODEL FOR EACH SERIES AND CHECK THE FITTED MODEL.
Oied MA Models
Euro
Pound
0.000216
2.180719
0.0293
'!$#
-0.03527
-1.87115
0.0614
-0.03906
-2.06828
'!""#
MA(1)
0.036145
1.917522
0.0553
MA(2)
-0.04918
-2.60871
0.0091
0.0387
Dollar
'!"#
'!"(#
Euro
-9.23493
-7.50935
Yen
.69 84
.8417
3.69079
4.458243
-.6 8
.7867
-1.9 6 9
4.1689
MA(9)
MA(1)
-10.3245
Correlogram of Residuals
.4
-7. 4
Correlogram of Residuals
^
Gollar
QUESTION 7: BUILD AN ARMA MODEL FOR EACH SERIES AND CHECK THE FITTED
MOD EL.
Pound
0.000214
1.998408
0.0458
!"#
-1.0633
-241.783
!$#
-0.98758
-225.352
'!"#
1.073462
350.7353
0.994895
326.4942
'!$#
AR(1)
-0.44664
-1.94152
0.0523
MA(1)
0.493893
2.206844
0.0274
-7.51275
Dollar
!"#
'!"#
Yen
-.46
. 6
-2.0827
2.427948
.74
.1
AR(1)
AR(2)
MA(1)
MA(2)
-9.23439
-8.64928
-1.784
-.9 1
1.799989
.94 66
-. 818
-. 18
8.4
.6
-7. 981
Euro
Correlogram of Residuals
^
Gollar
QUESTION 8: USE THE FITTED MODELS TO COMPUTE 1-DAY AHEAD AND 7 DAY AHEAD FORECASTS STARTING FROM CHOSEN DATE.
Model fied
AR
MA
ARMA
EURO
Equaion
AR Oorecas
.1196
.16769
.17998
MA Oorecas
.1196
.886
.41974
ARMA Oorecas
.1196
-.4 717
-. 748
Model fied
AR
MA
ARMA
POUND
Equaion
Model fied
AR
MA
ARMA
-8.791 6E-
9.71769E-
-. 97 6
E-9
-1.8 78E-6
4.76
DOLLAR
Equaion
Model fied
AR
MA
ARMA
. 91 187
. 91 187
.1 44
. 7964
-1.7 76E-
.18649
. 91 187
1.88 1E-
1.84 9 E-7
YEN
Equaion
. 96847
-.194E-
.199
. 96847
. 96847
-.1461E-
7. 9E-
.16
4. 8 9E-
Euro
Gollar
X
X
^
In all four log return series of exchange rates there is significant auto correlation in the squared return
series
P-value of Box-Ljung statistics in all four cases is zero (< confidence Interval level), there null
hypothesis is rejected, implying significant auto-correlation in squared residuals. (Null Hypothesis in
Box-Ljung is
hus arch-effec exiss in log-reurns of all four series
Pound : Ar(0)-Arch(8)
)
0.000242
1.998408
0.0458
!""#
-0.039388
-241.783
RESID(1)^2
M
)
2.29E-06
18.56635
M
0.07915
4.471953
RESID(2)^2
0.07534
6.209688
2.08E-05
28.46481
RESID(3)^2
0.069738
10.31774
G!"#*$
0.151156
8.348903
RESID(4)^2
0.068818
4.987416
G!$#*$
0.080704
5.096814
RESID(5)^2
0.165215
7.321421
G!%#*$
0.02391
1.794203
0.0728
RESID(6)^2
0.034923
2.203753
0.0275
G!&#*$
0.069687
3.914755
0.0001
RESID(7)^2
0.072234
4.711605
RESID(8)^2
0.028443
2.038036
0.0415
-7.59811
Dollar: Arma(1,1)-Arch(6)
-9.38981
Yen: AR(10)-Arch(6)
)
!"#
-.6 144
-4.28945
'!"#
.666
4.765568
)
AR(9)
-0.02876
-2.033008
0.0421
AR(10)
0.06106
4.290906
M
M
6.7E-7
31.33969
2.53E-06
20.23423
G!"#*$
(&"&+%%
",++"(-
RESID(1)^2
0.285686
14.57719
G!$#*$
. 8
11.69246
RESID(2)^2
0.062073
4.023763
0.0001
G!%#*$
.87
4.626349
RESID(3)^2
0.046324
3.244222
0.0012
G!&#*$
.14691
8.380267
RESID(4)^2
0.073219
4.094673
G!+#*$
. 8166
11.05533
RESID(5)^2
0.227713
9.200789
G!.#*$
.1 6
9.439572
RESID(6)^2
0.026184
2.387226
0.017
-9.6
-9.16182
Euro
Gollar
X
X
^
After fitting the appropriate arch model in all four series, auto-correlation in squared residuals were
significantly reduced
e also observed he auo-correlaion in residuals in arch as well as garch model fiing, i was ver
low. e have no repored he ACO/PACO graphs of residuals (due o fixed page limi of 15) afer
fiing arch and garch models, bu we checed ha residuals have low auo-correlaions afer fiing
hese models
11. BUILD A GARCH MODEL WITH GAUSSIAN INNOVATIONS FOR THE LOG RETURNS OF
EACH SERIES.
Oied Models
0.000196
AR(11)
-0.040662
3.30E-07
G!
"#*$
!
"#
8.98E-05
)
2.18153
Coefficient
0.0291
!"#
0.319475
0.02016
-2.0163
0.0438
'!"#
0.346465
7.53E-08
4.38998
3.82E-08
0.042265
0.004697
8.99829
0.947256
0.00598
158.263
G!
"#*$
!
"#
z-Statistic
Prob.
-2.02067
0.043
2.20269
0.027
8.64E-09
4.41732
0.035697
0.004125
8.65298
0.956934
0.004949
193.349
/
0
AR(10)
1|
0.044545
0.018031
0.15729
Variance equaion
POUND
Std.
Error
0.1581
)
2.470479
-7..&&
0.0135
Coefficient
!"#
-0.674
'!"#
0.71067
/
0
DOLLAR
Std.
Error
0.21094
0.0.1989
-9.&&
z-Statistic
Prob.
-3.19545
0.0014
3.57121
0.0004
Variance equaion
3.74E-07
7.67E-08
4.872609
4.01E-08
3.18E-09
12.6303
G!
"#*$
!
"#
0.06553
0.0049867
13.14309
0.17066
0.011241
15.182
0.926037
0.005787
160.0214
G!
"#*$
!
"#
!
$#
0.355055
0.074781
4.74794
.67
7.23862
.4899
,+$.
2&%"
^EN
1 . COMPARE THE FITTED ARCH AND GARCH MODELS. IS THERE ANY SIGNIFICANT
DIFFERENCE?
Arch and Garch models are the same as mentioned in previous questions
Series
Model
Arch
Euro Garch
Arch
Pound Garch
Arch
Dollar Garch
Arch
Yen Garch
X
X
Skewnes Kurtosis
s of
of
J-B
Innovations LL
AIC
SIC
DW
Residuals Residuals statistics
Normal
164.6 -7.98114 -7.8 7 . 8 .671 4.666 14.4817
Normal
17.9 -7.6448 -7.64 . 7 1.84 4. 64474 189.614
Normal
1 1.68 -9.8987 -9.778 1.9 9898 -. 4 .9748 1.9
Normal
1 64.7 -9.444 -9.4 98 1.98116 .781 4.8 4 6.174
Normal
11 1.9 -9.649 -9.17 1.9697 .976 9.79 4 6. 7
Normal
1 4. -9.4 69 -9.417 1.974 .6 6 9.8768 716.91
12835.55
-9.161822 -9.14274 1.957627 0.230323 4.935999 462.0336
Normal
12875.89
-9.19E+00 -9.18573 1.95892 0.20676 4.885251 434.603
Normal
Garch models across all log-return series have lower AIC, thus information loss is less, hence more
accurate. Same is observed from SIC.
Residuals have lower kurtosis in most cases, implying a closer resemblance to normal distribution
LOGRETURNPOUNG
LOGRETPOUNG(-1)
0.039956
(0.01999)
LOGRETPOUNG(-2)
-0.05259
(0.02003)
LOGRETPOUNG(-5)
-0.051952
(0.01993)
LOGRETPOUNG(10)
LOGRETPOUNG(11)
LOGRETPOUNG(12)
LOGRETGOLLAR(1)
LOGRETGOLLAR(2)
LOGRETGOLLAR(5)
LOGRETGOLLAR(10)
LOGRETGOLLAR(11)
LOGRETGOLLAR(12)
C
0.000507
(0.01998)
-0.006593
(0.02008)
-0.061694
(0.02004)
-0.033395
(0.03467)
0.009534
(0.03454)
0.106105
(0.03428)
0.035805
(0.03459)
0.025186
(0.03472)
-0.021052
(0.03459)
5.12E-05
(4.5E-05)
AIC
LOGRETURN GOLLAR
[
1.99888]
[2.62553]
[2.60714]
[
0.02536]
[0.32835]
[3.07858]
[0.96318]
[
0.27603]
[
3.09484]
[
1.03499]
[
0.72535]
[0.60867]
[
1.13445]
-0.053131
(0.01152)
-0.014392
(0.01154)
-0.022768
(0.01148)
0.031639
(0.01151)
0.013039
(0.01157)
-0.052003
(0.01155)
0.089575
(0.01998)
-0.016539
(0.01990)
0.030321
(0.01976)
0.070934
(0.01993)
-0.081744
(0.02001)
0.042007
(0.01993)
3.50E-05
(2.6E-05)
-9.23604
[4.61259]
[1.24687]
[1.98280]
[
2.74843]
[
1.12697]
[4.50329]
[
4.48342]
[0.83094]
[
1.53474]
[
3.55836]
[4.08543]
[
2.10767]
[
1.34565]
-10.3385
Cor(LOGRETPOUND,LOGRETDOLLAR(-i))
.04
.04
.02
.02
.00
.00
-.02
-.02
-.04
-.04
2
10
12
Cor(LOGRETDOLLAR,LOGRETPOUND(-i))
10
12
Cor(LOGRETDOLLAR,LOGRETDOLLAR(-i))
.04
.04
.02
.02
.00
.00
-.02
-.02
-.04
-.04
2
10
12
10
12
Pormaneau es
VAR Residual Portmanteau Tests for Autocorrelations
Null Hypothesis: no residual autocorrelations up to lag h
Date: 11/22/11
Time: 04:02
Lags
Q-Stat
Prob.
Adj Q-Stat
Prob.
df
0.036518
NA*
0.036531
NA*
NA*
0.144591
NA*
0.144682
NA*
NA*
6.437866
NA*
6.444711
NA*
NA*
11.43298
NA*
11.44697
NA*
NA*
11.47483
NA*
11.48890
NA*
NA*
12.34550
NA*
12.36144
NA*
NA*
14.94332
NA*
14.96578
NA*
NA*
18.10792
NA*
18.13945
NA*
NA*
23.95559
NA*
24.00599
NA*
NA*
10
24.07214
NA*
24.12296
NA*
NA*
11
24.10335
NA*
24.15429
NA*
NA*
12
24.21589
NA*
24.26732
NA*
NA*
13
26.61878
0.0000
26.68142
0.0000
14
46.56651
0.0000
46.72947
0.0000
15
58.19974
0.0000
58.42540
0.0000
12
16
61.44892
0.0000
61.69326
0.0000
16
17
64.63741
0.0000
64.90125
0.0000
20
18
74.83857
0.0000
75.16846
0.0000
24
19
77.52704
0.0000
77.87531
0.0000
28
20
95.79013
0.0000
96.26989
0.0000
32
LOGRETPOUND
LOGRETDOLLAR
Joint
4.101093
31.15588
46.74749
[ 0.128665]
[ 1.72e-07]
[ 1.72e-09]
7.274628
3.274298
8.832711
[ 0.026323]
[ 0.194534]
[ 0.065420]
12.48438
4.832243
13.97314
[ 0.001946]
[ 0.089267]
[ 0.007381]
1.218455
29.75408
30.77929
[ 0.543771]
[ 3.46e-07]
[ 3.40e-06]
0.535662
16.73656
21.40527
[ 0.765037]
[ 0.000232]
[ 0.000263]
12.31871
20.7693
27.7941
[ 0.002114]
[ 3.09e-05]
[ 1.37e-05]
Used Var lag exclusion test to remove lag-variables which are insignificant
Coefficien
|G
!"#
|G
!+#
|G
!-#
|G
!2#
|G
!"(#
|G
!""#
|G
!"$#
||
G!"#
||
G!+#
||
G!-#
||
G!2#
||
G!"(#
||
G!""#
||
G!"$#
|1|!
"#
|1|!
+#
|1|!
-#
|1|!
2#
|1|!
"(#
|1|!
""#
|1|!
"$#
|!
"#
|!
+#
|!
-#
|!
2#
|!
"(#
|!
""#
|!
"$#
sd error
-sa
-sa
logreurn_euro
-sa
-sa
0.109927
(0.02306)
[ 4.76728]
0.034197
(0.04020)
[ 0.85059]
0.055957
(0.04700)
[ 1.19044]
0.082576
(0.04147)
[ 1.99144]
0.035491
(0.02293)
[ 1.54757]
0.072476
(0.03999)
[ 1.81257]
0.027132
(0.04675)
[ 0.58038]
0.013980
(0.04124)
[ 0.33899]
0.043804
(0.02329)
[ 1.88092]
-0.02744
(0.04061)
[-0.67577]
0.126580
(0.04747)
[ 2.66628]
0.002826
(0.04188)
[ 0.06748]
-0.007847
(0.02340)
[-0.33529]
-0.044218
(0.04081)
[-1.08362]
-0.091701
(0.04771)
[-1.92209]
0.001724
(0.04209)
[ 0.04097]
0.072452
(0.02339)
[ 3.09697]
0.030910
(0.04079)
[ 0.75781]
0.138336
(0.04769)
[ 2.90073]
0.051406
(0.04207)
[ 1.22193]
-0.066565
(0.02340)
[-2.84416]
0.055850
(0.04081)
[ 1.36867]
-0.060307
(0.04771)
[-1.26405]
-0.007522
(0.04209)
[-0.17872]
0.013795
(0.02312)
[ 0.59663]
-0.040468
(0.04031)
[-1.00388]
0.023112
(0.04713)
[ 0.49037]
0.057909
(0.04158)
[ 1.39279]
-0.004898
(0.01477)
[-0.33164]
0.068462
(0.02575)
[ 2.65873]
-0.077996
(0.03011)
[-2.59069]
-0.001033
(0.02656)
[-0.03889]
0.003801
(0.01473)
[ 0.25806]
-0.055262
(0.02568)
[-2.15160]
-0.001833
(0.03003)
[-0.06103]
-0.079344
(0.02649)
[-2.99524]
0.017113
(0.01477)
[ 1.15876]
-0.007791
(0.02575)
[-0.30255]
-0.036771
(0.03011)
[-1.22138]
0.016465
(0.02656)
[ 0.61997]
0.010669
(0.01482)
[ 0.71975]
-0.053632
(0.02584)
[-2.07523]
-0.022205
(0.03022)
[-0.73489]
-0.007155
(0.02666)
[-0.26842]
0.043147
(0.01485)
[ 2.90564]
-0.009273
(0.02589)
[-0.35815]
0.078281
(0.03027)
[ 2.58604]
-0.052145
(0.02670)
[-1.95275]
0.011954
(0.01485)
[ 0.80495]
0.016971
(0.02589)
[ 0.65546]
0.012894
(0.03027)
[ 0.42594]
0.032794
(0.02670)
[ 1.22801]
-0.053129
(0.01476)
[-3.60067]
-0.088369
(0.02573)
[-3.43498]
-0.086688
(0.03008)
[-2.88206]
-0.076244
(0.02653)
[-2.87345]
-0.004062
(0.01147)
[-0.35422]
-0.042992
(0.02000)
[-2.15003]
0.018459
(0.02338)
[ 0.78956]
-0.00845
(0.02062)
[-0.40970]
-0.004786
(0.01138)
[-0.42071]
0.031476
(0.01983)
[ 1.58699]
-0.006253
(0.02319)
[-0.26966]
0.017430
(0.02046)
[ 0.85205]
-0.025795
(0.01144)
[-2.25534]
-0.017694
(0.01994)
[-0.88733]
-0.052927
(0.02331)
[-2.27009]
-0.01234
(0.02057)
[-0.60000]
-0.021922
(0.01145)
[-1.91404]
-0.004024
(0.01997)
[-0.20149]
-0.009786
(0.02335)
[-0.41913]
-0.005142
(0.02060)
[-0.24967]
0.004804
(0.01141)
[ 0.42115]
-0.000833
(0.01989)
[-0.04189]
0.020281
(0.02325)
[ 0.87223]
0.016761
(0.02051)
[ 0.81715]
-0.008505
(0.01139)
[-0.74639]
-0.014177
(0.01987)
[-0.71362]
-0.014772
(0.02323)
[-0.63598]
-0.005456
(0.02049)
[-0.26627]
0.021765
(0.01137)
[ 1.91370]
0.011380
(0.01983)
[ 0.57387]
0.021788
(0.02318)
[ 0.93979]
-0.024328
(0.02045)
[-1.18950]
-0.074863
(0.01467)
[-5.10210]
-0.046368
(0.02558)
[-1.81246]
0.034628
(0.02991)
[ 1.15769]
-0.019602
(0.02639)
[-0.74289]
-0.031945
(0.01464)
[-2.18212]
0.000710
(0.02552)
[ 0.02781]
-0.0494
(0.02984)
[-1.65537]
0.011491
(0.02633)
[ 0.43651]
-0.018819
(0.01463)
[-1.28642]
0.062892
(0.02551)
[ 2.46574]
0.012037
(0.02982)
[ 0.40363]
0.001941
(0.02631)
[ 0.07379]
0.009322
(0.01478)
[ 0.63089]
0.039625
(0.02576)
[ 1.53801]
0.022662
(0.03012)
[ 0.75232]
0.005211
(0.02657)
[ 0.19611]
-0.021707
(0.01476)
[-1.47073]
0.011405
(0.02573)
[ 0.44322]
-0.002838
(0.03009)
[-0.09432]
0.035363
(0.02654)
[ 1.33240]
0.000273
(0.01475)
[ 0.01852]
-0.032293
(0.02572)
[-1.25531]
0.010556
(0.03008)
[ 0.35096]
-0.056295
(0.02653)
[-2.12175]
-0.00183
(0.01483)
[-0.12337]
0.037474
(0.02586)
[ 1.44897]
0.031285
(0.03024)
[ 1.03462]
0.056858
(0.02667)
[ 2.13154]
9.87E-05
(6.0E-05)
[ 1.65278]
0.000110
(0.00010)
[ 1.05693]
0.000181
(0.00012)
[ 1.48417]
0.000213
(0.00011)
[ 1.98475]
/3|
I have used VAR exclusion test to remove insignificant lag variables from the estimated model. As seen from
the above test, all lag variables in the estimated model are significant at 8% level.
Autocorrelations w ith 2 Std.Err. Bounds
Cor(LOGRETDOLLAR,LOGRETDOLLAR(-i))
Cor(LOGRETDOLLAR,LOGRETPOUND(-i))
Cor(LOGRETDOLLAR,LOGRETEURO(-i))
Cor(LOGRETDOLLAR,LOGRETYEN(-i))
.06
.06
.06
.06
.04
.04
.04
.04
.02
.02
.02
.02
.00
.00
.00
.00
-.02
-.02
-.02
-.02
-.04
-.04
-.04
-.04
-.06
-.06
-.06
10
12
Cor(LOGRETPOUND,LOGRETDOLLAR(-i))
10
12
Cor(LOGRETPOUND,LOGRETPOUND(-i))
-.06
2
10
12
Cor(LOGRETPOUND,LOGRETEURO(-i))
.06
.06
.06
.06
.04
.04
.04
.04
.02
.02
.02
.02
.00
.00
.00
.00
-.02
-.02
-.02
-.02
-.04
-.04
-.04
-.04
-.06
-.06
-.06
10
12
Cor(LOGRETEURO,LOGRETDOLLAR(-i))
10
Cor(LOGRETEURO,LOGRETPOUND(-i))
10
12
Cor(LOGRETEURO,LOGRETEURO(-i))
.06
.06
.06
.04
.04
.04
.04
.02
.02
.02
.02
.00
.00
.00
.00
-.02
-.02
-.02
-.02
-.04
-.06
-.04
-.06
2
10
12
Cor(LOGRETYEN,LOGRETDOLLAR(-i))
10
12
10
12
Cor(LOGRETYEN,LOGRETEURO(-i))
.06
.06
.06
.04
.04
.04
.04
.02
.02
.02
.02
.00
.00
.00
.00
-.02
-.02
-.02
-.02
-.04
-.04
-.04
-.04
-.06
-.06
-.06
10
12
10
12
10
12
10
12
Cor(LOGRETYEN,LOGRETYEN(-i))
.06
12
-.06
2
Cor(LOGRETYEN,LOGRETPOUND(-i))
10
-.04
-.06
2
Cor(LOGRETEURO,LOGRETYEN(-i))
.06
-.04
-.06
2
12
Cor(LOGRETPOUND,LOGRETYEN(-i))
-.06
2
10
12
10
12
ep 1 : esing of Uni roos in Dollar and Pound eries (ADO es)
Null Hypothesis
AG
Result
Thus both dollar and pound series have a unit root and are non-stationary
59.93926
0.5854
102.3903
POUNG
-0.18573
0.007578
-24.50905
Adjusted R-squared
0.175879
Stats
R-squared
0.176172
Gurbin-Watson stat
0.008547
D-W is close to zero there we accept or fail to reject the null hypothesis of No Co-integration. Thus
no co-integration exist between dollar and pound series
Augmented
Gickey- -1.792
uller test statistic
Prob.*
0.3849
Thus we accept or fail to reject the null hypothesis implying that residuals have a unit root. Therefore
residuals from regression of dollar and pound are not stationary implying
4
5
6
0
4. Only if the variables are co-integrated, the residuals from the equilibrium regression
can be used to estimate the error correction model. In absence of co-integration VAR with equilibrium
correction is same as a normal VAR
17. ARE THE FOU R SERIES OF THE US DOLLAR, JAPANESE YEN, EURO AND UNITED
KINGDOM POUND EXCHANGE RATES VERSUS INDIAN RUPEES COINTEGRATED ? DISCUSS
THE IMPLICATIONS.
Trace
0.05
No. of CE(s)
Eigenvalue
Statistic
Critical Value
Prob.**
None
0.002564
17.5322
47.85613
0.9983
At most 1
0.002329
10.32968
29.79707
0.975
At most 2
0.001348
3.786496
15.49471
0.92
At most 3
1.49E-07
0.000417
3.841466
0.9856
As seen from he johansen co-inegraion es, here is no co-inegraion beween dollar, pound,
euro and en series a 0.05% level.
In absence of co-inegraion ECM i.e. error correcion model is no informaive, he model is same
as a normal var. Thus in absence of co-integration among the dollar, yen, euro and pound series,
building a four-variate auto-regression model with equilibrium correction will be useless and hence I
have not build a four-variate autoregressive model with equilibrium correction (Note: VAR with error
correction can be easily built using VEC option in eviews while estimating VAR, but in this case ECM is
useless).